Approximate letter grades:

• C: 22 to 40 (16 scores)
• B: 42 to 60 (9)
• A: 61 to 71 (8)

1. General

1.1

Would you argue that strict adherence to the ISO/OSI protocol layering is detrimental to the efficiency of communication?

Yes. By not delegating the authoriy to a lower level protocol, some processing time can be saved. This, however, impinges on the integrity of information hiding principle during the network software development process.

1.2

Why is ATM ( Asynchronuous Transfer Mode) network considered connection oriented while the Internet is considered connectionless?

ATM establishes virtual circuits for the duration of a cell transmission; the connection from the source to the destination host is at that point fixed.

1.3

Consider a network with physical connections existing between consecutive sites A, B, C, and D. The following picture illustrates timing of events in different types of switching networks, when a message is being sent from A to D through B and C.

1.3.1

Identify types of networks in (a), (b), and (c).

Circuit switching, message switching, and packet switching, respectively.

1.3.1

What is wrong with the part (a)?

The call accepting signal on the figure does not have the propagation delay (it should by sloping to the left).

1.4

Why isn't there an ICMP message that would allow a machine to inform the source that transmission errors are causing datagrams to arrive corrupted?

If transmission errors have caused a datagram to be corrupted, the correctness of its source address cannot be relied on.

2. Ethernet

2.1

Contrast the physical properties of Thinnet cable and the coaxial cable. What are the advantages and disadvantages of using one rather than the other as the Ethernet wire?

Thinnet is thinner and therefore more flexible, but also provides less protection against EM fields and also somewhat higher losses and smaller signal rate (capacity).

2.2

Given that a transmission in an Ethernet cable consists of a sequence of bits, what allows a host to separate these bits into frames?

The preamble field in an ethernet frame indicates the beginning of the frame and synchronizes the reception; the length field allows to recognize its size.

2.3

In the Ethernet, what aspects of the technology dictate the minimum frame size?

The maximum length of the cable -- through finite velocity of the signal.

2.4

What dictates the maximum frame size?

When the transmission error is detected, the whole frame has to be retransmitted. The probability of error is proportional to the number of bits in a frame. It pays to keep the frame size relatively small so that the frame retransmission does not happen too often.

3. Addressing

3.1

The text states that ``Each host on a TCP/IP internet is assigned a unique 32-bit internet address that is used in all communication with that host.'' (p.60) Is this statement true or not?

No, a multihomed host has several IP addresses, depending on the network through which it is accessed. Also, the address is not used during a local broadcast.

3.2

Consider a machine with two physical network connections and two IP addresses, I1 and I2. Is it possible for that machine to receive a datagram destined for I2 over the network specified in I1?

Yes, if a datagram destined to I2 is routed through the network specified in I1, then the machine can receive it to ``forward it'' to the other network.

3.3

What are applications in your computing environment that can benefit from IP broadcast? multicast?

Broadcasing messages to client hosts about an imminent server failure. Email to group aliases (a restricted set of hosts).

4. IP Datagram

4.1

Using the values FragmentLength, Offset, and MoreFragments associated with each incoming fragment, write an algorithm assembling the original datagram from a buffer Fgmts[1..n] of received fragments ( FragmentLength includes the 5 header words; assume that Offset gives directly the number of octets preceding the fragment data in the datagram.)

4.2

Why is the TimeToLive field in an IP header sometimes called hop count?

Each hop (routing through a next data link) causes the TimeToLive field to be decremented even if a full second has not elapsed.

5. Subnetting

5.1

Does it ever make sense to subnet a class C network?

Yes, for example for security reasons (one subnet within and the other outside the ``firewall''), or other functional differences.

5.2

Explain why a class B address using the third octet for the subnet address can not accomodate more than 254 subnets.

The address consisting of all ones corresponds to a local broadcast, while all zeros address refers to "this host".

5.3

What can happen if a packet arrives destined for a non-existent subnet?

It may be routed to the default router, from which it will come back to be delivered, until its time to live expires.

6. ATM

6.1

Why is the size of an ATM cell (53 octets) considered exceptional (while compared, for instance, with the Ethernet frame size)?

Normally, such a small frame size would imply too much bandwidth loss through the necessity of encapsulation.

6.2

An ATM switch transmits data at 155 Mbps. What is the duration of the pulse of light representing 1 bit in a fiber optic cable?

1/(155,000,000) = 1/.155 E-9 = 6.45 nsec

6.3

The speed of light is 300,000 km/s. What is the length of the fiber optic cable in which the time of a return trip equals the transmission time of one ATM cell?

An ATM cell has 53 octets, that is 424 bits that are transmitted in a 155 Mbps cable in

424 X 6.45ns = 2.7usec.

During half this time, the light can travel 400m, a quarter of a mile.

6.4

ATM allows a host to establish multiple virtual circuits to a given destination. What is the major advantage of doing so?

This allows the many cells into which a datagram is fragmented to be sent simultaneously, increasing the speed ot the transmission.